Pith. sign in

REVIEW 4 major objections 6 minor 45 references

Progressive Fine-to-Coarse Reconstruction for Accurate Low-Bit Post-Training Quantization in Vision Transformers

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A progressive fine-to-coarse reconstruction schedule for low-bit post-training quantization of vision transformers reports 75.61% Top-1 accuracy on a 3-bit ViT-B and brings 4-bit ViT-B within 1.35% of full precision.

desk verdict Plausible progressive reconstruction idea, but the headline results rest on a weak reproduced baseline and the SOTA claim overreaches. read the letter →

arxiv 2412.14633 v1 pith:XHBNFFKD submitted 2024-12-19 cs.CV cs.AI

Rui Ding , Liang Yong , Sihuan Zhao , Jing Nie , Lihui Chen , Haijun Liu , Xichuan Zhou This is my paper · ORCID
classification cs.CVcs.AI
keywords Post-TrainingQuantizationVisionTransformersReconstructionGranularityImageClassificationObjectDetectionInstanceSegmentationLow-bitModelCompression
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper is trying to establish that low-bit post-training quantization of vision transformers can be made much more accurate by reconstructing the quantized model at progressively coarser granularities, rather than at a single fixed granularity. The method starts from the two finest reconstruction units — the multi-head self-attention module and the MLP, each with its shortcut — and after optimizing them, combines pairs to form blocks, then pairs of blocks, and so on, with each coarser stage inheriting the optimized weights of the finer stage. This progressive fine-to-coarse reconstruction is paired with a two-stage optimization schedule (POS) that first quantizes only activations and then both weights and activations, which the paper says eases optimization. On ImageNet, the paper reports 75.61% Top-1 accuracy for a 3-bit quantized ViT-B and 4-bit ViT-B within 1.35% of full precision. The same method also improves object detection and instance segmentation on COCO, indicating the benefit is not limited to classification.

What carries the argument

The central object is the progressive fine-to-coarse reconstruction unit. The finest units are defined as $R^A_0(X_l)=X_l+\mathrm{MHSA}(\mathrm{LN}(X_l))$ and $R^P_0(Y_l)=Y_l+\mathrm{MLP}(\mathrm{LN}(Y_l))$, i.e., the attention and MLP modules with their shortcuts; these are optimized with mean-square-error losses, then combined into coarser units $R_g$ at granularity levels $g=1,\dots,G$ by the recursive composition in Eq. (16)-(17). The second mechanism is the Progressive Optimization Strategy (POS), a two-stage schedule: stage one quantizes only activations, keeps weights full-precision, and uses $G=1$; stage two quantizes both weights and activations with the full $G$, while diminishing learning rates and increasing iterations with $g$ according to Eq. (18)-(19). The combination of progressive inheritance of weights from finer to coarser granularities and the two-stage quantization schedule is what the paper identifies as the source of the accuracy gains.

What would settle it

Re-run the Table 3 ablation for 4-bit ViT-S under the paper's stated calibration data and hyperparameters, but replace the reproduced block-wise baseline with a published strong block-wise reconstruction such as I&S-ViT; if that baseline already achieves roughly 74.9% Top-1, then the 37.45% gain attributed to PFCR is an artifact of the weak baseline, and the per-block reconstruction-loss explosion reported for block-wise methods would not replicate.

Watch

Extended reading notes

Core claim

The central claim is that the reconstruction granularity used in post-training quantization should be progressively increased rather than fixed in advance. The paper shows that when finer units (attention and MLP with shortcuts) are reconstructed first, they provide a better initialization for reconstructing the coarser blocks built from them, and the coarser reconstruction in turn refines the parameters of the finer units, so that iterating this process suppresses accumulated reconstruction error. The paper substantiates this claim with per-block reconstruction-loss curves that stay flat under PFCR but explode under block-wise reconstruction, and with ImageNet accuracy gains that grow as the coarsest granularity level G increases (e.g., 3-bit ViT-S improves from 19.23% to 49.83% as G goes from 0 to 3 with fixed 250-iteration training). The reported headline result is 75.61% Top-1 accuracy for 3-bit ViT-B, outperforming prior state-of-the-art methods on the same setting.

Load-bearing premise

The load-bearing premise is that the reproduced block-wise reconstruction baseline used in the ablation (36.58% Top-1 for 4-bit ViT-S) is a fair and representative implementation of single-granularity reconstruction; if a properly tuned block-wise method already reaches about 74.9% on the same setting, the claimed 37.45% improvement from progressive fine-to-coarse reconstruction is inflated and the central superiority claim is not established.

Editorial extensions

If this is right

  • If the reported results hold, 3-bit and 4-bit ViTs can be deployed with accuracy close to full precision, making transformer-based vision models practical on memory-constrained devices.
  • The method extends to object detection and instance segmentation, so the reconstruction technique is not tied to the image-classification objective.
  • Reconstruction time drops by roughly a factor of seven compared with Outlier-Aware (about 19 minutes vs 130 minutes for 4-bit ViT-S) while improving accuracy, so low-bit PTQ becomes more practical.
  • The ablation shows that increasing the number of granularity levels G is itself the main driver of accuracy in low-bit settings, which indicates that the progressive schedule, not just more training, is responsible for the gains.
  • At 6-bit quantization the progressive schedule can slightly hurt accuracy due to overfitting on the small calibration set, so the method's best operating range is at 3-4 bits.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same fine-to-coarse schedule could likely be applied to other transformer-based architectures, including large language models, if the memory cost of the coarser granularity levels is reduced; the conclusion itself notes that direction as future work.
  • A validation-set-driven choice of the coarsest granularity level $G$ could avoid the mild 6-bit overfitting while preserving the 3-4 bit gains, since the paper fixes $G$ by a power-of-two formula rather than by data.
  • The two-stage POS suggests that activation-only quantization could serve as a universal warm start for any weight-and-activation PTQ method, not only for PFCR; this could be tested by pairing POS with existing single-granularity reconstruction methods.
  • The reported accuracy gains at 3 bits indicate that even finer reconstruction units (e.g., LayerNorm outputs or the post-softmax attention) might push useful PTQ down to 2 bits, where current ViT quantization methods typically collapse.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper proposes Progressive Fine-to-Coarse Reconstruction (PFCR) for post-training quantization of Vision Transformers. The method defines MHSA and MLP units with shortcuts as the finest reconstruction granularity, then progressively combines them into coarser blocks and reconstructs at each granularity level, with a two-stage Progressive Optimization Strategy (POS). Experiments on ImageNet and COCO report accuracy improvements over prior PTQ methods, particularly at 3-bit and 4-bit quantization (e.g., 75.61% top-1 for 3-bit ViT-B).

Significance. If the reported results hold, the paper offers a simple and practical improvement for low-bit PTQ of ViTs, with the progressive fine-to-coarse idea being conceptually clear and easy to integrate into existing reconstruction frameworks. The paper includes extensive ablations, efficiency comparisons, loss distribution analysis, and qualitative visualizations. The main weakness is that the quantitative support for the central claim currently rests on an ablation baseline that appears far weaker than the published block-wise method it is compared against, and the abstract overstates the SOTA claim.

major comments (4)
  1. [Section 4.4, Table 3] The block-wise baseline of 36.58% top-1 for 4-bit ViT-S is far below the 74.87% reported for the block-wise method I&S-ViT in Table 1 under the same model and bit-width. The paper states this baseline was reproduced 'under the same setting,' but with the default 300 iterations and 4e-5 learning rate it appears severely under-trained. Consequently, the claimed 37.45% improvement from PFCR over block-wise reconstruction is not a fair measure; against the I&S-ViT block-wise baseline, the improvement in Table 1 is only 1.44 points. This is load-bearing because the Introduction and Section 4.4 use this comparison to motivate the central claim that progressive fine-to-coarse reconstruction is superior to single-granularity block-wise reconstruction. The ablation should either use a properly tuned block-wise baseline or report the exact hyperparameter settings and explain why the reproduced baseline is so much weaker than the published number.
  2. [Abstract, Section 4.2, Table 1] The abstract states that the method 'achieves the best Top-1 accuracy among state-of-the-art methods,' but Table 1 shows it is second-best for 6-bit DeiT-S (79.34 vs Outlier-Aware 79.50) and 6-bit DeiT-B (81.65 vs I&S-ViT 81.68). The body of the paper acknowledges this in Section 4.2, so the abstract overstates the result. The claim should be qualified to 'best in most settings, particularly at 3-bit and 4-bit' or similar.
  3. [Eq. (17), Section 3.2] The recursive definition of R_g appears to contain an indexing error. For g>1 the formula reads R_g = R_{g-1}(R_{g-1}(X_{l-2^{g-1}+1})), but for a coarse unit spanning 2^g blocks ending at index l, the input should be X_{l-2^g+1}, not X_{l-2^{g-1}+1}. As written, the recurrence does not match the description of combining two adjacent R_{g-1} units and is not well-defined. Since Eq. (17) is the formal definition of the proposed reconstruction units, this needs to be corrected for reproducibility.
  4. [Tables 1 and 2] The main comparison tables report results from a single run without variance estimates. Given that calibration and reconstruction use randomly sampled 64 and 1024 images, and that several claimed improvements are as small as 0.07-0.3 points, reporting mean plus/minus standard deviation over multiple runs (as done for Figure 4) would be needed to substantiate the SOTA claims.
minor comments (6)
  1. [Section 3.2 title] The section title reads 'Progessive Fine-to-Course Reconstruction'; it should be 'Progressive Fine-to-Coarse Reconstruction.' There is also a typo in the same section: 'coaserest' should be 'coarsest.'
  2. [Eq. (15)] The notation uses the rounding operator ⌊·⌉ inconsistently with the floor operator ⌊·⌋ used in Eq. (3). For a power-of-two test, floor and rounding happen to coincide for the given examples, but the notation should be unified for clarity.
  3. [Figure 4] The caption states that 'We test three times to compute the mean and std,' but the figure does not show error bars or report the numerical mean and standard deviation values. Either add error bars or remove the statement.
  4. [Section 4.3, Table 2] The sentence 'APQ-ViT achieves the best box average precision with the Swin-S backbone' is immediately followed by 'Nevertheless, it suffers significant performance degeneration with the Swin-S backbone,' which is contradictory. The second clause should presumably refer to the Swin-T backbone, for which APQ-ViT is indeed poor.
  5. [Table 2 caption] The asterisk marker '∗' is used for RepQ-ViT results but is not explained in the caption. Please add a note that these are reproduced with the official code.
  6. [Section 4.4.2] The sentence 'boosting their classification performance by 38.67%. and 2.28% severally' has a misplaced period and should read 'by 38.67% and 2.28%, respectively.'

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: reported accuracies are direct measurements; the noted baseline-fairness caveats are evidence concerns, not circularity.

full rationale

The derivation chain in PFCR defines reconstruction units in Eqs. (8)-(9) and objectives in Eqs. (11)-(16) as MSE between quantized and full-precision outputs; no fitted constant is later relabeled as a prediction. The final Top-1 accuracies in Tables 1, 4, and 5 are direct empirical measurements on ImageNet, and hyperparameters such as lr0 and iter0 are tuned in the standard sense but do not determine the reported accuracy by construction. The paper does not rely on self-citations: I&S-ViT, Outlier-Aware, BRECQ, and other baselines are external prior works, and the two-stage POS is explicitly motivated by those prior observations rather than imported as an unexamined premise. The strongest caveat is baseline fairness: Table 3's reproduced 'Block-wise' baseline (36.58% at 4-bit ViT-S) is far below the same paper's Table 1 entry for the block-wise method I&S-ViT (74.87%), and the abstract's unconditional 'best Top-1 accuracy' claim is contradicted by Table 1 at 6-bit DeiT-S (79.34 vs 79.50 for Outlier-Aware) and 6-bit DeiT-B (81.65 vs 81.68 for I&S-ViT). These are correctness and evidence concerns about comparison strength, not circularity: the ablation numbers are measurements, not quantities defined in terms of the claimed result. No equation in the paper reduces to its own input by construction, so the circularity score is 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper contributes an optimization schedule but rests on a long list of hand-chosen hyperparameters, a domain assumption that MSE output matching transfers to task accuracy, and an unproven claim that progressive granularity is beneficial. The baseline fairness assumption is the most fragile.

free parameters (5)
  • Base learning rate lr0 = 4e-5 (ImageNet), 6e-7 (COCO)
    Chosen by hand; no sensitivity analysis; all reconstruction results depend on it.
  • Base iterations iter0 = 800 (3-bit), 300 (4-bit), 100 (6-bit) for ImageNet; 500 for COCO
    Chosen by hand to trade off time and accuracy; Figure 4 shows sensitivity but no selection criterion.
  • Granularity schedule decay factor = 0.2 in Eqs. 18 and 19
    Arbitrary constant scaling learning rate down and iterations up per granularity level.
  • POS Stage 1 coarsest level G = 1
    Set to 1 in stage 1 for smooth training; not derived, and results vary with this choice.
  • Reconstruction and calibration sample counts = 1024 reconstruction / 64 calibration images (ImageNet); 1 reconstruction sample (COCO)
    Hand-selected; affects performance and comparability with baselines.
assumptions (4)
  • domain assumption MSE between quantized and full-precision unit outputs is a good proxy for final task accuracy
    Central objective Eq. (16) assumes reducing output MSE transfers to Top-1 accuracy and detection metrics.
  • domain assumption Progressive fine-to-coarse reconstruction improves optimization because finer units provide better initialization
    Claimed in Sections 1 and 3.2 without formal proof; supported only by ablation.
  • standard math STE gradient approximation through the rounding function is valid for updating quantized ViT parameters
    Standard practice in PTQ; invoked in Section 3.1.2.
  • domain assumption Comparison baselines are reproduced under equivalent experimental settings
    The paper compares to published numbers and a self-reproduced block-wise baseline (36.58%) that is far below the cited I&S-ViT result (74.87%) for the same setting; this assumption is load-bearing.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Progressive Fine-to-Coarse Reconstruction for Accurate Low-Bit Post-Training Quantization in Vision Transformers." pith.science (2026). https://pith.science/paper/XHBNFFKD

@misc{pith2026241214633,
  author       = {Pith},
  title        = {Pith review of: Progressive Fine-to-Coarse Reconstruction for Accurate Low-Bit Post-Training Quantization in Vision Transformers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XHBNFFKD}},
  note         = {Machine review of arXiv:2412.14633}
}
read the original abstract

Due to its efficiency, Post-Training Quantization (PTQ) has been widely adopted for compressing Vision Transformers (ViTs). However, when quantized into low-bit representations, there is often a significant performance drop compared to their full-precision counterparts. To address this issue, reconstruction methods have been incorporated into the PTQ framework to improve performance in low-bit quantization settings. Nevertheless, existing related methods predefine the reconstruction granularity and seldom explore the progressive relationships between different reconstruction granularities, which leads to sub-optimal quantization results in ViTs. To this end, in this paper, we propose a Progressive Fine-to-Coarse Reconstruction (PFCR) method for accurate PTQ, which significantly improves the performance of low-bit quantized vision transformers. Specifically, we define multi-head self-attention and multi-layer perceptron modules along with their shortcuts as the finest reconstruction units. After reconstructing these two fine-grained units, we combine them to form coarser blocks and reconstruct them at a coarser granularity level. We iteratively perform this combination and reconstruction process, achieving progressive fine-to-coarse reconstruction. Additionally, we introduce a Progressive Optimization Strategy (POS) for PFCR to alleviate the difficulty of training, thereby further enhancing model performance. Experimental results on the ImageNet dataset demonstrate that our proposed method achieves the best Top-1 accuracy among state-of-the-art methods, particularly attaining 75.61% for 3-bit quantized ViT-B in PTQ. Besides, quantization results on the COCO dataset reveal the effectiveness and generalization of our proposed method on other computer vision tasks like object detection and instance segmentation.

Figures

Figures reproduced from arXiv: 2412.14633 by the authors.

Figure 1
Figure 1. Comparisons between the conventional reconstruc￾tion methods with the single and fixed granularity and our proposed progressive fine-to-coarse reconstruction method. Zhang, Yu, Wang and Gu (2021) first attempts to reconstruct the quantized model in a block-wise manner, reducing gener￾alization errors. PD-Quant Liu, Niu, Yuan, Yang, Wang and Liu (2023a) then improves the performance of block-wise reconstruction by in… view at source ↗
Figure 2
Figure 2. Illustration of the proposed PFCR. The left sub-figure shows the two finest reconstruction units with granularity level 𝑔 = 0. The right sub-figure demonstrates the iterative process of reconstruction and combination from fine granularity to coarse granularity progressively. 3.2. Progessive Fine-to-Course Reconstruction Reconstruction methods aim to restore performance by decreasing the output difference between qua… view at source ↗
Figure 3
Figure 3. Comparison of the reconstruction loss distribution in the 12-block 4-bit quantized ViT-S between conventional block-wise reconstruction and the proposed PFCR. (a) The overall loss distribution of 12 blocks in ViT-S. (b) The zoomed in loss distribution of the first 6 blocks in ViT-S. proposed PFCR achieves a significant 37.45% accuracy im￾provement under the same experimental setting. Furthermore, we explore the effe… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Quantization results of ViT-S in 3-bit, 4-bit, and 6- bit bit-width with increasing iteration number. (a) The Top-1 (%) accuracy and (b) the time cost (minutes) is adopted as the evaluation metric for performance and efficiency respectively. We test three times to comp…
Figure 5
Figure 5. Figure 5: Qualitative visualization results of 3-bit quantized ViT-B using I&S-ViT method and our proposed PFCR. We select 8 images from the validation set of ImageNet dataset and utilize the Grad-CAM Selvaraju et al. (2017) technique to derive the attention maps. as shown in […

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 40 canonical work pages

  1. [1]

    , author L \'e onard, N

    author Bengio, Y. , author L \'e onard, N. , author Courville, A. , year 2013 . title Estimating or propagating gradients through stochastic neurons for conditional computation . journal arXiv preprint arXiv:1308.3432

    arXiv 2013

  2. [2]

    , author Nagel, M

    author Bondarenko, Y. , author Nagel, M. , author Blankevoort, T. , year 2024 . title Quantizable transformers: Removing outliers by helping attention heads do nothing . journal Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) volume 36

    work page 2024

  3. [3]

    , author Vasconcelos, N

    author Cai, Z. , author Vasconcelos, N. , year 2018 . title Cascade r-cnn: Delving into high quality object detection , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp. pages 6154--6162

    work page 2018

  4. [4]

    , author Yuan, P

    author Cao, X. , author Yuan, P. , author Feng, B. , author Niu, K. , year 2022 . title Cf-detr: Coarse-to-fine transformers for end-to-end object detection , in: booktitle Proceedings of the AAAI conference on artificial intelligence , pp. pages 185--193

    work page 2022

  5. [5]

    , author Lin, M

    author Chen, M. , author Lin, M. , author Li, K. , author Shen, Y. , author Wu, Y. , author Chao, F. , author Ji, R. , year 2023 . title Cf-vit: A general coarse-to-fine method for vision transformer , in: booktitle Proceedings of the AAAI Conference on Artificial Intelligence , pp. pages 7042--7052

    work page 2023

  6. [6]

    , author Yang, Z

    author Chu, T. , author Yang, Z. , author Huang, X. , year 2024 . title Improving the post-training neural network quantization by prepositive feature quantization . journal IEEE Transactions on Circuits and Systems for Video Technology volume 34 , pages 3056--3060

    work page 2024

  7. [7]

    , author Dong, W

    author Deng, J. , author Dong, W. , author Socher, R. , author Li, L.J. , author Li, K. , author Fei-Fei, L. , year 2009 . title Imagenet: A large-scale hierarchical image database , in: booktitle Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , organization Ieee . pp. pages 248--255

    work page 2009

  8. [8]

    , author Qin, H

    author Ding, Y. , author Qin, H. , author Yan, Q. , author Chai, Z. , author Liu, J. , author Wei, X. , author Liu, X. , year 2022 . title Towards accurate post-training quantization for vision transformer , in: booktitle Proceedings of the 30th ACM international conference on multimedia , pp. pages 5380--5388

    work page 2022

Show all 45 references
  1. [9]

    , author Wang, P

    author Dong, B. , author Wang, P. , author Wang, F. , year 2023 . title Head-free lightweight semantic segmentation with linear transformer , in: booktitle Proceedings of the AAAI Conference on Artificial Intelligence , pp. pages 516--524

    2023

  2. [10]

    , author Beyer, L

    author Dosovitskiy, A. , author Beyer, L. , author Kolesnikov, A. , author Weissenborn, D. , author Zhai, X. , author Unterthiner, T. , author Dehghani, M. , author Minderer, M. , author Heigold, G. , author Gelly, S. , author Uszkoreit, J. , author Houlsby, N. , year 2021 . t...

    2021

  3. [11]

    , author McKinstry, J.L

    author Esser, S.K. , author McKinstry, J.L. , author Bablani, D. , author Appuswamy, R. , author Modha, D.S. , year 2020 . title Learned step size quantization , in: booktitle Proceedings of the International Conference on Learning Representations (ICLR)

    2020

  4. [12]

    , author Gkioxari, G

    author He, K. , author Gkioxari, G. , author Doll \'a r, P. , author Girshick, R. , year 2017 . title Mask r-cnn , in: booktitle Proceedings of the IEEE international conference on computer vision , pp. pages 2961--2969

    2017

  5. [13]

    , author Lou, Z

    author He, Y. , author Lou, Z. , author Zhang, L. , author Liu, J. , author Wu, W. , author Zhou, H. , author Zhuang, B. , year 2023 . title Bivit: Extremely compressed binary vision transformers , in: booktitle Proceedings of the IEEE/CVF International Conference on Computer ...

    2023

  6. [14]

    , author Sun, N

    author Jiang, Y. , author Sun, N. , author Xie, X. , author Yang, F. , author Li, T. , year 2024 . title Adfq-vit: Activation-distribution-friendly post-training quantization for vision transformers . journal arXiv preprint arXiv:2407.02763

    2024 arXiv

  7. [15]

    , author Ba, J

    author Kingma, D.P. , author Ba, J. , year 2014 . title Adam: A method for stochastic optimization . journal arXiv preprint arXiv:1412.6980

    2014 arXiv

  8. [16]

    , author Oikonomou, A

    author Kirtas, M. , author Oikonomou, A. , author Passalis, N. , author Mourgias-Alexandris, G. , author Moralis-Pegios, M. , author Pleros, N. , author Tefas, A. , year 2022 . title Quantization-aware training for low precision photonic neural networks . journal Neural Networ...

    2022

  9. [17]

    , author Li, X

    author Le, P.H.C. , author Li, X. , year 2023 . title Binaryvit: pushing binary vision transformers towards convolutional models , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. pages 4664--4673

    2023

  10. [18]

    , author Wang, D

    author Li, K. , author Wang, D. , author Liu, G. , author Zhu, W. , author Zhong, H. , author Wang, Q. , year 2024 . title Diagswin: A multi-scale vision transformer with diagonal-shaped windows for object detection and segmentation . journal Neural Networks volume 180 , pages 106653

    2024

  11. [19]

    , author Gong, R

    author Li, Y. , author Gong, R. , author Tan, X. , author Yang, Y. , author Hu, P. , author Zhang, Q. , author Yu, F. , author Wang, W. , author Gu, S. , year 2021 . title \ BRECQ \ : Pushing the limit of post-training quantization by block reconstruction , in: booktitle Proce...

    2021

  12. [20]

    , author Xiao, J

    author Li, Z. , author Xiao, J. , author Yang, L. , author Gu, Q. , year 2023 . title Repq-vit: Scale reparameterization for post-training quantization of vision transformers , in: booktitle Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pp. p...

    2023

  13. [21]

    , author Maire, M

    author Lin, T.Y. , author Maire, M. , author Belongie, S. , author Hays, J. , author Perona, P. , author Ramanan, D. , author Doll \'a r, P. , author Zitnick, C.L. , year 2014 . title Microsoft coco: Common objects in context , in: booktitle Computer Vision--ECCV 2014: 13th Eu...

    2014

  14. [22]

    , author Zhang, T

    author Lin, Y. , author Zhang, T. , author Sun, P. , author Li, Z. , author Zhou, S. , year 2022 . title Fq-vit: Post-training quantization for fully quantized vision transformer , in: booktitle Proceedings of the ThirtyFirst International Joint Conference on Artificial Intell...

    2022

  15. [23]

    , author Niu, L

    author Liu, J. , author Niu, L. , author Yuan, Z. , author Yang, D. , author Wang, X. , author Liu, W. , year 2023 a. title Pd-quant: Post-training quantization based on prediction difference metric , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and ...

    2023

  16. [24]

    , author Wang, L

    author Liu, X. , author Wang, L. , year 2024 . title Msrmnet: Multi-scale skip residual and multi-mixed features network for salient object detection . journal Neural Networks volume 173 , pages 106144

    2024

  17. [25]

    , author Yang, H

    author Liu, Y. , author Yang, H. , author Dong, Z. , author Keutzer, K. , author Du, L. , author Zhang, S. , year 2023 b. title Noisyquant: Noisy bias-enhanced post-training activation quantization for vision transformers , in: booktitle Proceedings of the IEEE/CVF Conference ...

    2023

  18. [26]

    , author Lin, Y

    author Liu, Z. , author Lin, Y. , author Cao, Y. , author Hu, H. , author Wei, Y. , author Zhang, Z. , author Lin, S. , author Guo, B. , year 2021 a. title Swin transformer: Hierarchical vision transformer using shifted windows , in: booktitle Proceedings of the IEEE/CVF Inter...

    2021

  19. [27]

    , author Shen, Z

    author Liu, Z. , author Shen, Z. , author Savvides, M. , author Cheng, K.T. , year 2020 . title Reactnet: Towards precise binary neural network with generalized activation functions , in: booktitle Proceedings of the European Conference on Computer Vision (ECCV) , organization...

    2020

  20. [28]

    , author Wang, Y

    author Liu, Z. , author Wang, Y. , author Han, K. , author Zhang, W. , author Ma, S. , author Gao, W. , year 2021 b. title Post-training quantization for vision transformer , in: booktitle Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) , pp. pag...

    2021

  21. [29]

    , author Li, H

    author Ma, Y. , author Li, H. , author Zheng, X. , author Ling, F. , author Xiao, X. , author Wang, R. , author Wen, S. , author Chao, F. , author Ji, R. , year 2024 . title Outlier-aware slicing for post-training quantization in vision transformer , in: booktitle Proceedings ...

    2024

  22. [30]

    , author Kim, D

    author Moon, J. , author Kim, D. , author Cheon, J. , author Ham, B. , year 2024 . title Instance-aware group quantization for vision transformers , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 16132--16141

    2024

  23. [31]

    , author Gross, S

    author Paszke, A. , author Gross, S. , author Massa, F. , author Lerer, A. , author Bradbury, J. , author Chanan, G. , author Killeen, T. , author Lin, Z. , author Gimelshein, N. , author Antiga, L. , et al., year 2019 . title Pytorch: An imperative style, high-performance dee...

    2019

  24. [32]

    , author Álvarez García, J.A

    author Rendón-Segador, F.J. , author Álvarez García, J.A. , author Salazar-González, J.L. , author Tommasi, T. , year 2023 . title Crimenet: Neural structured learning using vision transformer for violence detection . journal Neural Networks volume 161 , pages 318--329

    2023

  25. [33]

    , author Cogswell, M

    author Selvaraju, R.R. , author Cogswell, M. , author Das, A. , author Vedantam, R. , author Parikh, D. , author Batra, D. , year 2017 . title Grad-cam: Visual explanations from deep networks via gradient-based localization , in: booktitle Proceedings of the IEEE international...

    2017

  26. [34]

    , author Liu, G

    author Shang, Y. , author Liu, G. , author Kompella, R.R. , author Yan, Y. , year 2024 . title Enhancing post-training quantization calibration through contrastive learning , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. ...

    2024

  27. [35]

    , author Cao, L

    author Su, K. , author Cao, L. , author Zhao, B. , author Li, N. , author Wu, D. , author Han, X. , author Liu, Y. , year 2024 . title Dctvit: Discrete cosine transform meet vision transformers . journal Neural Networks volume 172 , pages 106139

    2024

  28. [36]

    , author Cord, M

    author Touvron, H. , author Cord, M. , author Douze, M. , author Massa, F. , author Sablayrolles, A. , author J \'e gou, H. , year 2021 . title Training data-efficient image transformers & distillation through attention , in: booktitle Proceedings of the International Conferen...

    2021

  29. [37]

    , author Du, B

    author Wang, Y. , author Du, B. , author Wang, W. , author Xu, C. , year 2024 . title Multi-tailed vision transformer for efficient inference . journal Neural Networks volume 174 , pages 106235

    2024

  30. [38]

    , author Wang, C

    author Wang, Z. , author Wang, C. , author Xu, X. , author Zhou, J. , author Lu, J. , year 2022 . title Quantformer: Learning extremely low-precision vision transformers . journal IEEE Transactions on Pattern Analysis and Machine Intelligence volume 45 , pages 8813--8826

    2022

  31. [39]

    , author Gong, R

    author Wei, X. , author Gong, R. , author Li, Y. , author Liu, X. , author Yu, F. , year 2022 . title QD rop: Randomly dropping quantization for extremely low-bit post-training quantization , in: booktitle International Conference on Learning Representations , pp. pages 1--19

    2022

  32. [40]

    , year 2019

    author Wightman, R. , year 2019 . title Pytorch image models . howpublished https://github.com/rwightman/pytorch-image-models . :10.5281/zenodo.4414861

    2019 doi

  33. [41]

    , author Li, Z

    author Xu, K. , author Li, Z. , author Wang, S. , author Zhang, X. , year 2024 . title Ptmq: Post-training multi-bit quantization of neural networks , in: booktitle Proceedings of the AAAI Conference on Artificial Intelligence , pp. pages 16193--16201

    2024

  34. [42]

    , author Xue, C

    author Yuan, Z. , author Xue, C. , author Chen, Y. , author Wu, Q. , author Sun, G. , year 2022 . title Ptq4vit: Post-training quantization for vision transformers with twin uniform quantization , in: booktitle Proceedings of the European Conference on Computer Vision (ECCV) ,...

    2022

  35. [43]

    , author Yu, L

    author Zhang, N. , author Yu, L. , author Zhang, D. , author Wu, W. , author Tian, S. , author Kang, X. , author Li, M. , year 2024 . title Ct-net: Asymmetric compound branch transformer for medical image segmentation . journal Neural Networks volume 170 , pages 298--311

    2024

  36. [44]

    , author Hu, J

    author Zhong, Y. , author Hu, J. , author Lin, M. , author Chen, M. , author Ji, R. , year 2023 . title I&s-vit: An inclusive & stable method for pushing the limit of post-training vits quantization . journal arXiv preprint arXiv:2311.10126

    2023

  37. [45]

    , author Su, W

    author Zhu, X. , author Su, W. , author Lu, L. , author Li, B. , author Wang, X. , author Dai, J. , year 2021 . title Deformable \ detr \ : Deformable transformers for end-to-end object detection , in: booktitle Proceedings of the International Conference on Learning Represent...

    2021

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.